Workshops Archive

Descriptive Execution of HPC Applications and Workflows


Workshop: AgenticAI4HPC’26: The First International Workshop on Agentic AI for HPC

Authors: Vanessa Sochat (Lawrence Livermore National Laboratory (LLNL)) and Daniel Milroy (Lawrence Livermore National Laboratory)

Abstract: The means to execute and orchestrate software in high performance computing (HPC) has changed from human-written code to descriptive prose. We assess the extent to which an agentic framework can optimize and run an HPC scaling study with a low-latency network in Amazon Web Services, translate job specifications between workload managers, and design and run workflows. In job translation experiments, agents convert 982 of 1,000 specifications between Slurm and Flux, all validated by production Flux, matching human-written performance in all but one case. In the scaling experiments, agents deploy and optimize applications, demonstrating scaling through the largest size of nodes. In the bioscience workflow, the agent reproduces an expert-written variant-calling pipeline almost exactly -agreeing with the reference call set in 18 of 19 evaluable runs - through many distinct yet functionally equivalent implementations. This information is invaluable for enabling multi-cluster application deployment, with scheduling and transformation handled by agents.


Back to AgenticAI4HPC’26: The First International Workshop on Agentic AI for HPC Archive Listing

Back to Full Workshop Archive Listing